Priority Execution of Short-Time Jobs (How to Start Batch Job Execution Earlier)

When there are many running jobs or jobs waiting for execution in the queue, it takes a long time until execution starts even after submitting a job.
The status of jobs can be checked with the pjstat or pjstat2 command, and the status of resource group usage can be checked with the show_rsc command.

The show_rsc command does not show the nodes allocated for fixed node use, but when a node is not used by a fixed user, another user’s short-time job can be executed.
The status of these free nodes is not displayed in show_rsc.

Some compute nodes can run only short-time jobs when no reservation jobs are running, for the sake of leaving room to run reservation jobs.
The status of these free nodes is displayed by show_rsc.
(It is included in the a-batch/a-inter and b-batch/b-inter with simplex mode.)

Conditions for Executable Short-Time Jobs

All of the following conditions must be met:

• It specifies a shared resource group (a-batch, a-inter, b-batch, or b-inter).
• It is a node-occupying job (does not specify vnode-core or gpu).
• The elapsed request time (elapse) is less than 1 hour (fixed-use node) or less than 30 minutes (reserved-job node).

If the compute nodes are busy, please consider using a short-time job.
(Note that jobs cannot be executed if they are waiting due to Concurrent Core Usage Limits.)